With the rapid growth of photovoltaic power generation systems, fire incidents within the system have progressively increased. The lack of thorough studies on the temperature properties of direct current (DC) arc faults has resulted in an unclear ignition mechanism, significantly increasing the fire risk associated with such faults. Hence, this work presents a proposed experimental scheme for detecting photovoltaic DC series arc faults (SAFs) and the corresponding detection standards. Additionally, the temperature characteristics of the DC arc fault are further analyzed. The magnetohydrodynamic (MHD) arc fault simulation model is developed to investigate the temperature-related aspects of photovoltaic DC arc faults. Finally, our experimental validation confirms the precision of the model in simulating arc temperature. It is verified that the research presented in this paper can provide a good explanation for the rise time of DC arc temperature and the characteristic distribution of arc distance. This study elucidates the impact mechanism of line current, power supply voltage, and arc gap size on arc temperature in a photovoltaic system. Additionally, it proposes an evaluation method for assessing the arc fault ignition risk level. This method is essential for safeguarding against arc fault ignition risk in photovoltaic DC series cells.

Low-voltage AC power lines are prone to arc faults, and an arc current presents as a random and complicated signal. The amplitude of the line current remains relatively unchanged during the occurrence of series arcs, hence complicating the detection of series arc faults. In this work, we developed a low-voltage series arc fault test platform to analyze the digital features of low-voltage series arc currents and the morphology of arc combustion, as the current model fails to capture the high-frequency and randomness of arc currents. An analysis of the physical causes and influencing factors of the random distribution of AC arc zero-crossing times was conducted. A time-domain simulation model for arc fault currents was developed by enhancing the time constant of the Cassie arc model, while the high-frequency features of arc currents were simulated using a segmented noise model. The measured arc current data were utilized to validate the model through the analysis of the zero-crossing time distribution of arc current, the correlation coefficient of the arc current frequency-domain signal, and the similarity of the time-domain waveforms. When comparing the similarity of the simulated waveforms of the arc model presented in this research and those of other traditional arc models, it was found that the suggested model effectively characterizes the time-/frequency-domain features of low-voltage AC series arc fault currents. The suggested model enhances the features of randomness in low-voltage AC series arc faults and is important in extracting essential aspects and reliably recognizing low-voltage series arc faults.

Electrical fires are frequently caused by low-voltage AC series arc faults, which can result in significant injuries and property damage. The installation of arc-fault detection devices is mandated or recommended in many regions and countries across the world, yet the current devices’ detection accuracy is insufficient to completely eliminate the risk posed by arc faults. The method based on artificial intelligence is a solution with high detection accuracy, but the AI model is a ‘black box’. When a misjudgment occurs, the cause of the model error cannot be found fundamentally, and the modification and light weight of the model also presents significant difficulties when using the approach. Given the aforementioned issues, this research proposes a novel lightweight low-voltage AC arc-fault detection method based on the explainability approach. By applying the attention mechanism approach and performing a visual analysis, the contribution of arc features to model detection is determined. Model input data optimization and model structure simplification are achieved at the same time as increased model detection accuracy. Ultimately, an experimental prototype for arc-fault detection is designed and validated. Test results demonstrate the effectiveness of the method by demonstrating that the lightweight model maintains 99.69% detection accuracy, even after optimizing the input data by 80% and reducing the model parameters by 51.52%.

With the deep penetration of renewable energy and power electronic equipment, the overcurrent protection of an intelligent miniature circuit breaker faces new challenges. The electronic controller of an intelligent miniature circuit breaker is typically powered by the bus current rather than the phase voltage to ensure a robust overcurrent protection response under all conditions, including severe short-circuit faults. So, the performance of the current transformer serving as an energy harvesting unit and the corresponding direct current to direct current convention circuit is one of the critical issues due to the limited volume of an intelligent miniature circuit breaker. In this research, a finite element model of a current transformer for an intelligent miniature circuit breaker is constructed by COMSOL to evaluate the impact of the core material, the core size, and the number of coil turns on the energy harvesting capability of the current transformer. Meanwhile, the relationship between the output of the power supply and its design parameters is investigated by circuit simulation. As a result, a novel type of current transformer is proposed based on well-designed parameters. Finally, experimental tests have been conducted to verify the hysteresis characteristics, output characteristics, and energy harvesting effect. The results demonstrate that the hysteresis properties of the transformer align with the simulation results. The power supply can work with a minimum current of 8 amperes, which is 23.08% better than before.

Low-voltage AC series arc faults are a leading cause of electrical fires, posing significant risks to life and property. While artificial intelligence-based detection methods have achieved high accuracy, they often suffer from limited interpretability and are typically tailored to specific loads, thus struggling to adapt to the diverse and dynamic load conditions in residential environments. To address these limitations, this paper proposes a novel interpretable arc fault detection model based on prototype learning with a hybrid attention mechanism. Specifically, we design a Tri-Domain Dynamic Attention (TDDA) module that integrates time-domain, frequency-domain, and temporal derivative information, and embed it into a Convolutional Neural Network (CNN) for enhanced feature extraction. Visual prototypes are constructed from sample characteristics, forming a tri-domain arc fault prototype set. A dedicated non-arc prototype set is further introduced to refine the decision boundary and improve accuracy. The proposed model is validated through comprehensive experiments and hardware implementation on a dedicated test platform. Results demonstrate that our model achieves an accuracy of 99.65%, maintains over 99% accuracy across various single-load conditions, and exhibits high detection performance under complex multi-load scenarios.

With the growing global focus on sustainable development and climate change mitigation, promoting the low carbonization of energy systems has become an inevitable trend. Power load monitoring is crucial to achieving efficient power management, and load identification is the key link. The traditional load identification method has the problem of low accuracy. It is assumed that the technique of fusing harmonic features through color coding can improve the accuracy of load identification. In this paper, the load’s instantaneous reactive power, power factor and current sequence distribution characteristics are used as the mapping characteristics of the R, G and B channels of the two-dimensional V–I trajectory color image of the load using color coding technology. The harmonic amplitude characteristics are integrated to construct the mixed-color image of the load. The void residual shrinkage neural network is selected as the classification training model. The advantages and disadvantages of two residual shrinkage construction units, RSBU-CS and RSBU-CW, are analyzed. A single-load identification model with three RSBU-CWs is built. Different datasets verify the performance of the model. Compared with the test results of the ordinary color image dataset, the accuracy of the mixed-color image dataset is above 98%, and the accuracy of load identification is improved.